OVERVIEW
How might generative AI improve the lives of the elderly?
For the Google x Design @ Columbia UX Challenge, my team and I chose the challenge prompt which asked us to consider how generative AI might be applied to solve problems currently facing elderly users. I was personally excited by this prompt because it made me think about how to distill the possibilities of generative AI into a simple and intuitive interface, accessible by users of all ages.
Teammates: Jesslyn Andriono, Ainsley Cunningham, Katherine Ju, Alekhya Maram
Time frame: February – March 2024
SECONDARY
RESEARCH
Designing for the elderly is about more than just making UI bigger.
In our secondary research, we found evidence to suggest that accessible design involves greater thought than just making cosmetic changes to an existing UI.
Having too many features or tools presented all at once can be overwhelming (Salman et al., 2018)
Touch commands can be difficult to execute for those with mobility issues (Dodd et al., 2017)
We live in an ‘aging world’ where more and more people need accessible interfaces (Pollack, 2005)
PRIMARY
RESEARCH
Our initial exploratory interviews revealed that many users, beyond just the elderly, are unaware of Gemini and what it can do.
We conducted 5 interviews with Google users aged 65+ and asked them about their experiences using technology products. We were particularly interested in which Google products they used most frequently, and how much they knew about generative AI and Google’s Gemini.
Google Search and GMail were the most commonly used Google products among the users we surveyed
Many users expressed that they enjoyed features like Autofill in Google Search, which anticipated what they were searching for
Very few users were aware that Google’s Gemini existed or of how generative AI tools could add value to their lives
Affinity map
PROBLEM STATEMENT
How might we make Gemini’s features more accessible and intuitive for elderly users?
For our solution, we decided to integrate Gemini into Google Search, a product that the elderly users we surveyed were already very familiar with.
Our solution needed to show users…
What Gemini is
How Gemini can provide value
How Gemini and Google are related
IDEATION
Low-fidelity iterations
Concept 1: Gemini Search Assistant
This was my first try at integrating Gemini into Google Search’s autofill. I chose to focus on autofill, because it was one of the features that elderly users seemed to enjoy and use the most in the existing Google Search product. As users type, new prompts would show up that related to their evolving search query, each of which would be clickable.
I also tried thinking about how Gemini might be marketed in order to frame it as more friendly and personable.
Based on feedback from our Google mentors, we decided to revise this first concept and simplify our integration. Our mentors mentioned that having Gemini generate so many suggestions as the user typed might create even more informational overload.
Concept 2: Gemini Concierge
For my second concept, I imagined a customizable concierge of Gemini actions for the new tab page. The services or widgets added to a user’s dock would anticipate and suggest questions in those domains that would be relevant to that user.
The aim of this concept was to help elderly users understand exactly what Gemini is and how it might provide value to them in their daily lives. By categorizing questions into ‘packages’ or ‘services,’ it might take away the overwhelming quality of generative AI. Rather than framing it as, “Ask Gemini anything,” this would create a more compartmentalized and contextualized search experience.
After receiving another round of feedback from our Google mentors, we decided to continue working with this concept and attempt to combine elements of it with our previous concept.
Interlude: Launch of Google’s own integration
While we were in the midst of generating low-fidelity prototypes, Google launched a Gemini integration in the Google Search mobile app experience. In the Google app, users could now toggle between Google and Gemini at the top of their screen. We decided to work with this toggle concept for our own integration, which would introduce similar functionality to the desktop search experience, beyond just the Google mobile app.
We believed that although the toggle concept seemed effective, the visual experience of toggling between Google and Gemini on the mobile app was too jarring because the interfaces for these two services looked very different from each other.
This might be confusing for not only elderly users but also for any user who was not yet aware of Gemini. Google’s mobile integration improved the accessibility of Gemini, but did not yet demonstrate the potential value of Gemini to new users.
Mid-fidelity iteration
Gemini Tools
Abstracting the earlier Gemini Concierge concept, we distilled Gemini’s primary functions into 5 tools: Write, Edit, Ideate, Explain, and Create. These tools would be accessible via a minimal button on the new tab page of Google Search.
Prompt templates
In case the blank Gemini prompt box seemed at all intimidating to a new user, we included a series of Gemini templates for each tool that could populate the prompt box with variables such as text fields, checkboxes, and dropdowns to help the user tweak their prompt.
UI as conceptual reinforcement
Leaning into the imagery of the word ‘Gemini,’ we laid out the Gemini Tools in the shape of a constellation. In switching between the Google and Gemini tabs, we wanted the interface to stay as similar as possible, to reinforce the notion of the two products being like the Gemini ‘twins.’
USABILITY
TESTING
We returned to the users to see whether our Gemini integration was useful and intuitive.
Users unfamiliar with AI were able to familiarize themselves with Gemini’s functionalities through our new Gemini integration. They especially enjoyed the new autofill interface. We also learned, however, that single word descriptors of the tools like Write or Edit were too broad and vague for many elderly users to grasp.
We also sought feedback from our mentors at Google, who suggested that our constellation UI might not be native enough to Google. By their suggestion, we decided to change this out for a card-based design in the final prototype.
FINAL
PROTOTYPE
CONCLUSION
Next steps and takeaways
Due to the time constraints of this 8-week design challenge, we were unable to fully flesh out all of the features we were interested in designing. But, if we had more time, here are some of the additional projects we would have wanted to explore.
Funnels to Google products, Google ads, and Google content creators
The power of AI lies beyond mere text or image generation. We believe that integrating the ability to ask Gemini to perform actions within other Google apps would greatly increase the value of Gemini for users, especially those with mobility or vision issues. Speaking to the needs of Google’s advertisers and content creators, embedding links in Gemini results might also be a useful step.
Onboarding tutorials for new users
From some of the feedback we received during usability testing, we learned that new users might not know what the Gemini Tools button does, or they may be confused as to how exactly to use it. To remedy this, we imagine that a series of simple dialog pop-ups could appear for first time users as a kind of tutorial on how to use this feature.
Final presentation at Google
In late March, my team and I presented our process and final prototype at Google in NYC.
